11256995

System and Method for Prediction of Protein-Ligand Bioactivity Using Point-Cloud Machine Learning

PublishedFebruary 22, 2022
Assigneenot available in USPTO data we have
Technical Abstract

Patent Claims
20 claims

Legal claims defining the scope of protection, as filed with the USPTO.

1

1. A system for prediction of protein-ligand bioactivity comprising: a computing device comprising a memory and a processor; a point-cloud based bioactivity module comprising a first plurality of programming instructions stored in the memory and operating on the processor, wherein the first plurality of programming instructions causes the computing device to: receive a molecular structure file, wherein the molecular structure file comprises molecular and structural information about at least one protein and one ligand; generate a graph-based neural network of the protein, wherein edges of the graph-based neural network of the protein are determined using the molecular structure file; generate a graph-based neural network of the ligand, wherein edges of the graph-based neural network of the ligand are determined using the molecular structure file; concatenate a set of vectors from both the graph-based neural network of the protein and the graph-based neural network of the ligand; perform restricted-cross-attention learning on the concatenated vectors; generate a single feature vector from the restricted-cross-attention learning; use the single feature vector in a feed-forward neural network to produce one or more outputs selected from the group consisting of active or inactive classification, crystalline-structure similarity, and regression analysis; and use the one or more outputs to produce one or more bioactivity predictions about one or more protein-ligand pairs.

2

2. The system of claim 1 , wherein docking simulations are performed on the protein and the ligand from the molecular structure file if the protein and the ligand are coupled.

3

3. The system of claim 1 , wherein the vector from the graph-based neural network of the protein comprises atoms within 4 angstroms of the ligand if the protein and the ligand are coupled.

4

4. The system of claim 1 , wherein the vector from the graph-based neural network of the protein comprises atoms of a binding pocket if the protein and the ligand are decoupled.

5

5. The system of claim 1 , wherein the crystalline-structure similarity prediction is used to determine the legitimacy of the one or more bioactivity predictions.

6

6. The system of claim 1 , wherein the graph-based neural network of the protein and the graph-based neural network of the ligand are based on one or more transformer convolution classifiers.

7

7. The system of claim 1 , wherein the one or more outputs are used with a loss function to produce a trained model for bioactivity prediction.

8

8. The system of claim 1 , wherein the machine learning model is a transformer convolution classifier.

9

9. The system of claim 1 , further comprising an output of a three-dimensional visualization of the one or more protein-ligand pairs.

10

10. The system of claim 9 , wherein the three-dimensional visualization comprises features of molecular interaction properties.

11

11. A method for prediction of protein-ligand bioactivity comprising the steps of: receiving a molecular structure file, wherein the molecular structure file comprises molecular and structural information about at least one protein and one ligand; generating a graph-based neural network of the protein, wherein edges of the graph-based neural network of the protein are determined using the molecular structure file; generating a graph-based neural network of the ligand, wherein edges of the graph-based neural network of the ligand are determined using the molecular structure file; concatenating a set of vectors from both the graph-based neural network of the protein and the graph-based neural network of the ligand; performing restricted-cross-attention learning on the concatenated vectors; generating a single feature vector from the restricted-cross-attention learning; using the single feature vector in a feed-forward neural network to produce one or more outputs selected from the group consisting of active or inactive classification, crystalline-structure similarity, and regression analysis; and using the one or more outputs to produce one or more bioactivity predictions about one or more protein-ligand pairs.

12

12. The method of claim 11 , wherein docking simulations are performed on the protein and the ligand from the molecular structure file if the protein and the ligand are coupled.

13

13. The method of claim 11 , wherein the vector from the graph-based neural network of the protein comprises atoms within 4 angstroms of the ligand if the protein and the ligand are coupled.

14

14. The method of claim 11 , wherein the vector from the graph-based neural network of the protein comprises atoms of a binding pocket if the protein and the ligand are decoupled.

15

15. The method of claim 11 , wherein the crystalline-structure similarity prediction is used to determine the legitimacy of the one or more bioactivity predictions.

16

16. The method of claim 11 , wherein the graph-based neural network of the protein and the graph-based neural network of the ligand are based on one or more transformer convolution classifiers.

17

17. The method of claim 11 , wherein the one or more outputs are used with a loss function to produce a trained model for bioactivity prediction.

18

18. The method of claim 11 , wherein the machine learning model is a transformer convolution classifier.

19

19. The method of claim 11 , further comprising an output of a three-dimensional visualization of the one or more protein-ligand pairs.

20

20. The method of claim 19 , wherein the three-dimensional visualization comprises features of molecular interaction properties.

Patent Metadata

Filing Date

Unknown

Publication Date

February 22, 2022

Inventors

Alwin Bucher
Alvaro Prat
Orestis Bastas
Aurimas Pabrinkis
Gintautas Kamuntavicius
Mikhail Demtchenko
Sam Christian Macer
Zeyu Yang
Cooper Stergis Jamieson
Zygimantas Jocys
Roy Tal
Charles Dazler Knuff

Want to explore more patents?

Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.

Citation & reuse

Analysis on this page is generated by Patentable — an AI-powered patent intelligence platform. AI-generated summaries, explanations, and analysis may be reused with attribution and a visible link back to the canonical URL below. Patent abstracts and claims are USPTO public domain.

Cite as: Patentable. “SYSTEM AND METHOD FOR PREDICTION OF PROTEIN-LIGAND BIOACTIVITY USING POINT-CLOUD MACHINE LEARNING” (11256995). https://patentable.app/patents/11256995

© 2026 Patentable. All rights reserved.

Patentable is a research and drafting-assistant tool, not a law firm, and does not provide legal advice. Documents we generate are drafts for review by a licensed patent attorney.